Process Optimization under Parameter Uncertainty Conditions in CCU Process

Abstract

In order to produce methanol using CO2 and H2, several processes about the related catalytic reaction has been suggested and simulated. There may be systematic uncertainties like parameters of reaction kinetics, so a modelling method considering the parametric uncertainty is proposed and can give more informative data compared to the conventional modelling methods. However, the resulting distributional model requires large computational burdens due to the iterative calculations for convergence. To solve the problem, generalized extreme value distribution (GEVD) and neural network (NN) modellings are utilized. The formation parameters of GEVD are fitted by NN and as a result distributional reactor model in an explicit formulation is proposed. Compared to the shallow structured neural network for learning the formulation parameters, the deep neural network shows improved performance especially adjacent to the boundary layers of process inputs. As its explicit and distributional formulation, the proposed model is expected to be utilized for real-time stochastic model based approaches in optimization and control with reduced computational load.